I'll say the quiet bit first. I'm a big Claude user. I made the switch over from GPT a while back, and now Cowork, Projects, all the md files I keep, they're basically a system I've built to support the work: writing, design, flow, heavy information analysis, the sheer lifting of it. If you saw my setup you might think I know what I'm doing. Honestly, no. I'm not an expert. It's new, and it's been trial, a lot of error, and more trial.
So when people ask if AI is going to replace instructional designers, I don't think that's the interesting question. I've been doing this 15+ years, PGCHE-qualified, and I've already lived through a few "this changes everything" moments. Authoring tools changed how we build. LMS platforms changed how we deliver. AI's just the next one. The question worth asking isn't whether it replaces us. It's what it's actually good for, and what it isn't.
The bullet points era
I remember exactly when this started properly, back in the early ChatGPT days. Clients would send me reams of bullet points, pages of them, thinking I had no idea what an LLM even was. "I've written this course," they'd say, and they had, no judgement, no stigma, they'd just had some help. Then they'd ask me to use my expertise and turn thirty pages of that into an actual course, and what came back read like a dissertation.
That's the point worth making properly, because it still holds now. One prompt does not work. I'll say it again. ONE PROMPT. DOES NOT WORK. Getting something usable out of AI is a series of iterations, with real parameters, and that's exactly where the heavy lifting still happens, whatever the tool. Back then it was a simple prompt like "write this in prose, don't use bullet points" that turned a dump of bullets into an actual flow. From there you'd shape the sentence structure, the rhythm, make it read like a person wrote it. LLMs have got a lot better since. But the tell-tale signs are still there if you know what you're looking at, and the underlying truth hasn't moved. It's never one prompt and done.
Fast and cheap was never the answer
Here's a comparison I keep coming back to. If you build a loft extension fast and cheap, it doesn't stand up. Well, it might stand up, but you know what I mean! Cheap and value for money are not the same thing. Cheap cuts corners. Value for money is efficiency, and I believe AI can absolutely create value for money. That's a different conversation. But fast and cheap, on its own, still isn't the answer, whatever's doing the building.
Same goes for the integrity panic in higher education. Is GenAI an issue for assessment? Sure. But it's no different to calculators, or Google, or the internet before it. It's here to stay. Don't panic about the tool. Use the time it gives you back for the conversations that actually matter, why a student wrote something the way they did, what a People team is really trying to achieve.
The stigma is the wrong fight
There's a stigma around this that I think is completely backwards, the idea that using AI to help you write something means you cheated. If you're not using AI to help you, you're the one falling behind. Writing with AI is fine. Where it goes wrong is when someone writes with it, hands it to a client as entirely their own, and never rereads it, never reshapes it, never actually makes it theirs.
For what it's worth, this post is written with AI. I don't see why that needs hiding. The interesting part was never whether AI touched it.
Where it actually saves me time
I use Storyline a lot. I'm not a coder, and I won't pretend to be, but if you stay within triggers and variables, AI can build you a proper spec for something complicated fast. Say you've got a gamified activity where learners are spending coins, working out every variable, every trigger, and testing it until it behaves used to eat days. Now it's an hour or two. Same with branching scenarios, working out how many branches you actually need and what route they take. That's real time back, not theoretical time back.
Where I actually land
I'm not worried about AI. I'm worried about the volume of drivel it's letting people put out, unedited, unread, un-thought-about.
You can spot it a mile off. Pages of identical sentence structure, no rhythm, no flair, nothing that sounds like it passed through an actual person on the way out. Reams of information because it was easy to generate, not because anyone decided it needed to exist. And every time a learner sits through something like that, it costs the whole idea of online learning a bit more trust.
That's the bit worth losing sleep over. Not the tool. The output nobody bothered to own.
If you want to talk through where AI genuinely fits in your organisation's learning content, and where it really doesn't, that's a conversation I'm always up for. Email me or grab 30 minutes, or just take a look at what else we're doing at exceleratelearning.co.uk.
